
EHRs consist of several information management
strategies for various purposes. There have been numerous
data breaches cases reported which thereby confirms that the
security of these reports has to be increased further.
There have been many cases where EHRs have failed in
the area of security. Listed below are some incidents which
proves this point. [8]
Patient records were leaked in University of Michigan
Medical Center patient records were disclosed on the internet
the password.
In Florida, a public health employee was able to access
thousands of names of HIV infected patients and he sent it to
many local news stations
The University of Washington Medical network facility
was accessed by unauthorized personnel and patient data was
compromised
Medical centres in New York and Holland and their
weakness in protecting patient data was exposed
A group of researchers from University of Minnesota
inadvertently released names of kidney donors to the public.
Exposing a patient’s private information to the general
public can have disastrous effects on the reputation and
mental health of the patient. This sort of information leak can
also lead to financial losses to the individual involved.
The cases mentioned clearly indicates the unintended
problem that comes up with the usage of EHRs. The Health
care model proposed in this paper deals with many of these
problems and provides meaningful solutions. [9]
In our proposed model we provide solution to almost
all of the problems mentioned above.
VI. OUR PROPOSED MODEL
In our proposed model we have used both Blockchain
technology and Machine Learning Algorithms to provide a
better solution in terms of security. By using Machine
Learning, we provide additional features which can be the
base of ideas for further implementations on this subject.
Machine Learning is based on the concept of
centralization of data, while Blockchain technology uses
decentralization of data to provide high security. In this paper
we have tried to project our model which showcases how we
can use both these for this particular application.
A. Implementation of Blockchain
In this model, dual Blockchain structure is used, the first
part grants access to health data and is built using the
Hyperledger Fabric. The second part of the structure works
on Ethereum and performs all application and services.
Medical information is very sensitive and personal so a
closed Blockchain such as Hyperledger Fabric helps in
retaining necessary privacy required.
Majorly blockchains are classified as public Blockchains
and permissioned Blockchains.
This can be explained by considering the example of a
user wants who to sell a book to person with some rebate and
does not intend to tell about this to general public, the seller
then can employ permissioned Blockchain to hide the
information about the offer from the public. This model uses
a double encryption mechanism on a permission-based
Blockchain. The security that is provided by this model
which uses Blockchain is beyond and far more advanced than
any other centralized security system being used.
Furthermore, the patient’s data is made inaccessible and
unalterable. The Blockchain acts as a pointer, and provides
the direction to the location of the stored data, meaning that
anyone attempting to access patient data will be denied.
The health data is secured between the patient and the
authorized doctor. When the authorised doctor adds
additional information to the patients record history, the
system will automatically update it. Only those clinicians
who have authorized access can view the updates. None of
the doctors are given permanent authorization, the access for
the doctor ends when the patient wants so that the doctor can
no longer update the record or access it.
This is vital in scenarios where there is a need to change
the doctor in charge, therefore, with the help of Blockchain
the information transfer will be easy and secure. Issues
associated with the transfer of information by medical
institute employees is completely eliminated and there will
be no more data leaks probable in this transaction process and
also overseas transactions of information can be cost
effective as compared to the conventional techniques to do
the same.
In emergency situations when the patient is unconscious
and unable to provide any sort of input on his health, it would
be vital to have access to the patients’ health records. This
information while performing lifesaving surgeries as history
of past medications and illnesses are crucial before
performing any sort of major surgery.
B. Implementation of Machine Learning
There are two steps in building a new Machine learning
model. The first step is to take in the dataset and adjust the
model weights to increase the accuracy of the model. The
second step is testing the Machine learning model on
independent or new datasets for the accuracy of the model and
thus validate the model and prevent overfitting of the model.
An over fitted model is very good at a given dataset but is bad
at hypothesizing for the given problem. The procedure of
Machine learning in our proposed model has been depicted in
Fig.5.
Once a Machine learning model has been trained using
Supervised Learning it can be used to do various tasks such
as prediction, classification on the untrained dataset. In the
proposed model, we use the “Bag of Words” algorithm
which will extract only the required dataset and ignore the
various other things like the Name, Age, Address and other
personal details of the patient to maintain the privacy.
Supervised Learning builds a mathematical model of the
given dataset that contains both input and output data.
Supervised Learning can be used for classification and
regression.
Steps for Supervised Learning:
First, the category of training set data is selected
The set of input data and its complimentary outputs are
collected.
The accuracy of the trained function depends on how the
input dataset is represented. Typically, the input dataset is
Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020
978-93-5406-901-7 © 2020 IEEE
Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.